Normalization of Brain Morphology After Surgery in Sagittal Craniosynostosis

Abstract

Object

Nonsyndromic craniosynostosis (NSC) is associated with significant learning disability later in life. Surgical reconstruction is performed before 1 year of age to correct the cranial vault morphology and to allow for normalized brain growth with the goal of improving cognitive function. Yet, no studies have assessed to what extent this primary endpoint of surgery in allowing for normalized brain growth is actually achieved. Recent advances in magnetic resonance (MR) imaging have allowed for automated methods of objectively assessing subtle and pronounced brain morphology differences. We utilized one such technique, Deformation-based morphometry (DBM) Jacobian mapping, to determine how previously treated sagittal NSC (sNSC) adolescents differ significantly in brain anatomy compared to healthy matched controls up to 11.5 years after surgery.

Methods

Eight adolescent patients with sNSC, previously treated via whole vault cranioplasty at a mean age of 7 months, and 8 age- and IQ-matched (mean age for both groups = 12.3 years) control subjects without craniosynostosis, underwent functional MR imaging at 3 Tesla. Statistically significant group tissue volume differences were assessed using DBM; a whole brain technique that estimates morphologic differences between two groups at each voxel (p < 0.01). Group-wise Jacobian volume maps were generated using a spacing of 1.5 mm and a resolution of 1.5 × 1.5 × 1.5 mm³.

Results

There were no significant areas of volume reduction or expansion in any brain area(s) in sNSC adolescents compared to controls at p < 0.01. At the more liberal threshold of p < 0.05 two areas of brain expansion extending anteroposteriorly in the right temporo-occipital and left fronto-parietal regions appeared in sNSC compared to controls.

Conclusion

Compared to previous reports on untreated infants with sNSC, adolescents with sNSC in this small cohort had few areas of brain dysmorphology many years after surgery. This suggests that comprehensive cranioplasty performed at an early age offers substantial brain normalization by adolescence but also that some effects of vault constriction may still persist after treatment. Specifically, few areas of expansion in fronto-parietal and temporo-occipital regions may persist. Overall, data from this small cohort support the primary goal of surgery in allowing for more normalized brain growth. Larger samples, and correlating degree of normalization with cognitive performance in NSC, are warranted.

Introduction

Craniosynostosis (CSO) results from the premature pathologic fusion of the calvarial suture(s). It is an uncommon, but not rare, condition of infancy occurring in 1 in 2,000 to 1 in 2,500 live births.^{9, 48} Fusion of the suture(s) results in abnormal calvarial and concomitant brain morphology as the tissue is forced to expand into areas where there is lack of restriction. Isolated fusion of the sagittal suture, sagittal nonsyndromic craniosynostosis (sNSC), represents the most common form of CSO, comprising 40–60% of all isolated cases³⁴, and results in the hallmark head shape deformity, scaphocephaly. If left untreated, the deformity progresses; therefore surgical release of the vault is performed in the first year of life to allow for both normalized head shape as well as more normalized brain growth.

Today, recent reports show that up to 50% of children with sNSC grow to have some form of cognitive disability, despite having IQs in the normal range.^{5, 37, 56} The most common issues reported in sNSC children include those involving attention and behavior (executive function), language, and visuospatial processing.^{33, 37} The impetus for early intervention is to improve or attenuate these perceived cognitive deficits, with the assumption that surgery allows for a more normalized brain growth pattern.

Yet, to date, no studies have been done to assess whether this primary endpoint of surgery is reached in older children with sNSC, i.e., whether the vault remodeling in infancy does in fact ultimately lead to a normal brain morphology. We aimed to address this question using a novel advanced magnetic resonance imaging (MR) technique, Deformation-Based Morphometry, Jacobian determinant mapping, to quantify where and how the brains of sNSC adolescents, up to 11.5 years after whole vault reconstruction, differ in anatomy in relation to a group of age- and IQ-matched healthy adolescent controls.

Methods

Subjects

This was a prospective cohort study performed with approval from the Yale School of Medicine Human Investigations Committee (protocol number 1004006656). We studied 8 adolescents (mean age = 12.3 years) with sNSC (treated by Drs. Persing and Duncan) via whole vault cranioplasty at Yale-New Haven Hospital and 8 control children (mean age = 12.3 years) without craniosynostosis. The children with sNSC did not exhibit signs of syndromic craniosynostosis (specifically extracranial skeletal manifestations), and both groups of children were without cranial prosthesis, mental retardation, known neurological or psychiatric disorder, history of traumatic head injury or intracranial hemorrhage. As illustrated in Table 1, the two groups were group-matched by age, gender, race, and handedness, as well as by performance intelligence quotient (PIQ) and verbal intelligence quotient (VIQ) as measured by the Wechsler Intelligence Scale of Children 3rd edition (WISC-III).⁵⁸

Table: 1.

Characteristics of sagittal synostosis subjects and controls

	Corrected Sagittal Synostosis Children	Control Children	P
N	8	8	NS
Age, years (s.d.)	12.3 (1.8)	12.3 (1.6)	NS
Gender	6 M 2 F	7 M 1 F	NS
Race	7 W, 1 AA	7 W, 1 AA	NS
Handedness	8 Right	8 Right	NS
Age of Operation, months (s.d.)	7 (2)		NS
WISC-III Testing			NS
Performance IQ (s.d.)	111 (15)	115 (10)	NS
Verbal IQ (s.d.)	100 (16)	120 (16)	NS

Scan Protocol and Image Acquisition

All of the MRI scans were conducted on a single 3 Tesla Siemens (Erlangen, Germany) Tim Trio MR system with a 32-coil polarized head coil. Images were obtained using a sagittal 3D integrated parallel acquisition technique (iPAT) with TR = 1900ms, TE = 2.96ms, FA = 9°, matrix size = 256 X 256, slice thickness = 1 mm and 160 contiguous slices.

Registration

To facilitate better registration, FSL’s brain extraction tool was used to remove the skull and meninges off each subject’s 3D anatomical image. Any remaining skull or meninges were removed manually. Five non-linear registrations were computed within the Yale BioImageSuite software package⁴³ between the individual brain extracted 3D anatomical image and a commonly used 3D reference image (the Colin Brain, Holmes et al, 1998)²⁶ in Montreal Neurological Institute space (Evans et al, 1993)¹⁹ using the intensity-only component of the method reported in Papademetris and colleagues.⁴⁴ The first registration was done with a control spacing of 15 mm. It was then used as the starting point for the second registration, which was done with a control spacing of 10 mm. The three subsequent registrations each used the previous registration as the starting point and continued to get more refined with a control spacing of 5 mm, 2 mm and finally 1.5 mm. All subjects’ registrations were visually inspected to ensure accuracy.

Deformation-Based Morphometry and Jacobian Mapping

Jacobians, or Jacobian determinants, are measures of regional tissue volume change when an individual’s brain is warped to common space during non-linear registration.³ Jacobians are partial first-order derivatives of a vector-valued function, such as that used to warp brains onto a common template in MR imaging.^{12, 25} Jacobian values indicate how much expanding or contracting a given subject’s brain undergoes during alignment, and thus can be used to quantify the anatomic variability between groups. Jacobian values (J) can be produced for any given brain voxel in the reference coordinate system (x, y, z) to describe the point-wise volume change induced by the transformation for a subject:

$$J(x,y,z)=\left|\begin{array}{ccc}1+\partial {\text{u}}_{\text{x}}/\partial \text{x}& \partial {\text{u}}_{\text{x}}/\partial \text{y}& \partial {\text{u}}_{\text{x}}/\partial \text{z}\\ \partial {\text{u}}_{\text{y}}/\partial \text{x}& 1+\partial {\text{u}}_{\text{y}}/\partial \text{y}& \partial {\text{u}}_{\text{y}}/\partial \text{z}\\ \partial {\text{u}}_{\text{z}}/\partial \text{x}& \partial {\text{u}}_{\text{z}}/{\partial }_{\text{y}}& 1+\partial {\text{u}}_{\text{z}}/\partial \text{z}\end{array}\right|$$

Jacobian values greater than 1 indicate tissue contraction, values less than 1 indicate tissue expansion, values of 0 indicate folding, and infinite values indicate tearing. The production of these determinants allows one to generate parametric maps of local tissue anatomic variability – contraction and expansion -- for each subject. The individual maps generated in a patient group can thus be compared to the individual maps in a control group. This comparison reveals whether there are brain areas that require more expansion or contraction during registration, which in turn informs us of the magnitude of differences in gross morphology anywhere in the brain in one group relative to the other.¹⁴ This provides an objective, automated analysis with high regional sensitivity to anatomic differences between two groups throughout the entire brain (deep and superficial structures) and gives an excellent, quantitative assessment of both subtle and pronounced morphology differences.^{21, 49, 55, 57}

Jacobian Analysis

The non-linear registrations computed within BioImageSuite were used to produce Jacobian maps of local expansion/contraction where each voxel has a value representing the local volume change required to map that voxel of an individual participant’s brain to the reference MNI brain. The Jacobian maps were generated using a spacing of 1.5 mm and a resolution of 1.05 × 1.05 × 1.05 mm³. This process was done for all participants in both sNSC and control groups using a spacing of 1.5 mm and resolution of 1.05 mm³ (Figure 1a–b). Average maps were then generated using the individual Jacobian maps for each group (Figure 2a–b). Finally, the sNSC and control average Jacobian maps were compared against each other. To measure the smoothness of the data, AFNI’s 3dFWHmx program was used on the residuals of the t-tests and found a smoothness of 6.57 × 7.22 × 6.87 mm. This smoothness was imputed into AFNI’s AlphaSim program to determine the proper cluster for multiple comparison correction which resulted in a cluster of 1775 mm³ for a threshold of p < 0.01 and 7425 mm³ for a threshold of p < 0.05. When comparing the sNSC and control groups, voxels that had a t value greater than 2.974 (p < 0.01) and were part of a spatially continuous cluster size of 1775 mm³ were considered to be significantly different. Also included were the more liberal data where voxels had a threshold of t value greater than 2.144 (p < 0.05) and were part of a spatially contiguous cluster size of 7425 mm³.

Figure 1.

a-b. Examples of individual Jacobian maps generated for both (a) a control and (b) sNSC subject. Blue represents areas of brain volume contraction. Red represents areas of brain volume expansion. Spacing of 1.5 mm and resolution of 1.05 mm³.

Figure 2.

a-b. Average Jacobian group maps generated for (a) controls and (b) sNSC subjects. Blue represents areas of brain volume contraction. Red represents areas of brain volume expansion. Spacing of 1.5 mm and resolution of 1.05 mm³. Areas of localized brain volume expansion and contraction are qualitatively similar for both groups.

Results

As illustrated in Figure 3, direct statistical comparison of Jacobian parametric maps between the two groups revealed no areas of pronounced or localized group brain differences at the predetermined threshold of p < 0.01; specifically, there were no significant areas of volume reduction or expansion in any brain area(s) in sNSC adolescents compared to controls at p < 0.01. Only at the more liberal threshold of p < 0.05 did areas of localized brain expansion appear in sNSC compared to controls. These areas of brain expansion extended anteroposteriorly in the right tempo-occipital and left fronto-parietal regions (Brodmann areas [BA] 41, 39, 22, 19 and 6, 4, 3, 2, 1, respectively).

Figure 3.

a-b. Voxel-wise Jacobian comparison of average maps between sNSC and controls reveals (a) no pronounced or localized differences in brain morphology (localized brain volume contraction or expansion) between sNSC adolescents and controls (p < 0.01), (b) but two areas of brain expansion extending anteroposteriorly in the right temporal and left fronto-parietal regions at p < 0.05. Blue indicates areas of brain constriction in sNSC vs controls. Red indicates areas of brain expansion in sNSC vs controls. Grey indicates no difference. Spacing 1.5 mm; resolution 1.05 mm³.

Discussion

A critical question in sagittal nonsyndromic craniosynostosis (sNSC) is that of cognitive function. The most recent data strongly suggest that sNSC infants can be identified to have significantly impaired cognitive, motor, and/or language development prior to treatment.^{7, 10, 11, 13, 32, 35, 50} Starr and colleagues recently found sNSC infants have up to a 2.0 increased odds of delayed mental, psychomotor, and/or language development compared to their healthy, typically developing counterparts.⁵¹ Even after treatment, up to 50% of adolescents with sNSC continue on to experience problems in attention and behavior, language, and visuospatial processing despite having a normal IQ.^{5, 37, 56} The goal of early surgery to provide excellent cosmetic outcomes through vault reconstruction has been well established.^{15, 23, 38, 39, 42} However, the more important goal of allowing for more normalized brain development and improved cognitive performance later in life is an area of great import.

Recent studies have offered evidence of the value of early vault release in preventing or abating cognitive deficit later in life. A large, well-powered, and appropriately designed study using the most up-to date neurocognitive assessments revealed that earlier vault release, performed before 6 months of age, is associated with improved intellectual and developmental quotient outcomes later in life.⁴⁵ Furthermore, another recent study demonstrated that a more comprehensive remodeling technique confers a significantly better intellectual and developmental outcome compared to the widely employed modified strip craniectomy.²⁴ Modified strip craniectomy has gained popularity recently as it is associated with less operative time, less blood loss, and can be performed at an earlier age^{5, 4, 8, 31} but the aforementioned findings indicate that long-term outcomes on functional performance may not be equivalent. While further study is needed, these studies seem to align and indicate that maximum neurocognitive benefit is achieved by early comprehensive vault remodeling.

The coupling of advanced functional MRI (fMRI) and diffusion tensor imaging (DTI) provided novel insight into understanding the aforementioned cognitive deficit in these children. Until recently, the few existing basic clinical MRI studies have failed to identify any overt and consistent brain structure malformations in sNSC.^{18, 29} Recently, fMRI demonstrated that sNSC children have altered functional brain connectivity in regions responsible for the cognitive functions observed to be delayed in the sNSC population.⁶ Specifically, the prefrontal cortex, which is responsible for executive function, and the left lateral parietal cortex, which is responsible for language and visuospatial processing, have significantly altered connectivity in sNSC children compared to controls. Significant connectivity differences involving the default mode network (DMN), which is the primary network at rest but also needed to perform well in school-like tasks²², is also altered in sNSC children as compared to controls. Finally, DTI revealed trends towards altered white matter microstructure in sNSC children. Taken together, these results point to a more insidious process that results not in overt brain malformation in NSC but rather more subtle functional and microstructural disorganization that may, in part or whole, underlie the cognitive disabilities so commonly observed in these children.

An important goal, then, of early surgical intervention, is to abate or improve the cognitive dysfunction in these children by hopefully allowing for more normalized brain growth after vault constriction release, and attenuate whatever insidious process may be taking place. Still, no studies have evaluated whether this very basic assumption and goal of surgery is achieved.

This study, the first of our knowledge to evaluate brain normalization in previously treated sNSC adolescents, indicates that substantial brain normalization is achieved when early comprehensive surgical intervention is done in infancy. Although there have been many functional, cosmetic, and cognitive outcomes reports on this condition,^{5, 6, 23, 37, 38, 56} none addressed the basic question of whether the primary goal of surgery, that is, to allow for more normalized brain growth and development, is actually achieved. Brain dysmorphology in sNSC infants who have not undergone surgery has been well documented. Before surgery, sNSC infants demonstrate significant lengthening of the brain anteroposteriorialy, lengthening between the posterior horns of the lateral ventricles and occipital poles, mediolateral expansion of structures in the anterior frontal region, and mediolateral constriction notable on the left side.² Alrdridge and colleagues also examined brain morphology changes following surgery in a small cohort of subjects, studying the ratios of linear distances between 32 identifiable cortical and subcortical brain structures.¹ As expected, ratios of these distances differed significantly in untreated sNSC infants as compared to controls. Following surgery, the brains of the sNSC infants assumed a much more normal globular shape, but ratios of distances between structures were still deviated from normal, and overall brain shape reorganized to a morphology somewhere between normal and where it was pre-operatively. Thus, Alrdridge and colleagues helped to answer the very important question of whether surgery impacts the trajectory of brain growth, revealing that certain aspects of brain morphology soon after intervention did, in some way, revert to a more normalized pattern. However, limitations of the study were that analysis was only performed at 1 year after surgery (post-surgery age range at time of assessment: 1.4 – 2.4 years), and it was not based on objective, whole brain assessment to determine morphology differences throughout the entire brain. These factors are significant because brain volume doubles by 6 months, triples by 2 1/2 years, continues to grow rapidly until age 3, and then begins to decelerate until age 6, when the cranial vault reaches 90% of its adult size.^{30, 46} The brain also continues to grow steadily in size across childhood, and experiences another developmental spurt in adolescence, not reaching its full adult weight until the late teens or early 20s.^{30, 46, 52} During this time, synaptogenesis, pruning, and myelination are fervently active and contribute to the overall shape, size, and connectivity that is reached.⁵⁴ This current study, which relies on more sensitive, objective, and whole brain analysis and examines adolescent outcomes of infant surgical patients (with a mean age of 12.3 years), is more appropriate for quantifying true final morphology differences since the brains of our subjects have had sufficiently more time to grow and potentially normalize (11–11.5 years after surgery).

Our findings demonstrate that the brains of these patients, at least in this small cohort, do, in fact, appear to continue along a more normalized trajectory following surgery, to the point that they reach substantial normalization by early adolescence. Compared to the extensive morphology differences seen in the brains of sNSC infants,^{1, 2} our adolescent cohort showed few areas of significant differences in brain morphology. Specifically, there were no areas of significant brain expansion or contraction between sNSC and controls at the predetermined threshold of p < 0.01. The use of the more liberal threshold of p < 0.05 demonstrated two areas of brain expansion extending anteroposteriology in the right temporo-occopital and left fronto-parietal regions in sNSC adolescents compared to controls. However, these two areas of brain expansion conceivably represent residual areas of the fronto-parietal and temporo-occopital brain expansion seen in sNSC infants before treatment.^{1, 2} This argues that while the majority of the brain might achieve normalized morphology by adolescence (i.e. absence of mediolateral contraction, extensive diffuse anteroposterior expansion, and dysmorphology between deep brain structures), some regions may not. The implications of these changes on cognition and long-term outcomes are yet to be determined. Small changes in functional organization over time can lead to cognitive deficit, and attempting to allow for a more normalized morphology or organization through surgery may be what drives potentially better cognitive outcomes,^{40, 41} which is illustrated again by higher intelligence and developmental status through earlier and more comprehensive skull remodeling.^{24, 45} There is mounting evidence that significant deviations in loco-regional brain shape are associated with neurocognitive and executive function disorders, including ADHD and Autism.^{17, 20, 57} It is also known that the two brain regions found to be dysmorphic in our small cohort of previously treated sNSC subjects are responsible for language and visuospatial as well as verbal and spatial processing.^{16, 27, 47, 53} Because of this, it may become even more imperative that early comprehensive brain shape normalization be achieved, and the lasting morphology effects be thoroughly studied.

Moving forward, larger, prospective longitudinal samples of subjects, coupled with extensive neurocognitive testing, are needed. We know from neurodevelopmental studies that there exist critical periods in early brain development, and that prolonged disruption of functional organization during these periods can lead to permanent deficits.^{28, 36} As discussed above, constriction of the vault appears to play some role in creating deficit during this critical period in sNSC infants, as revealed not only by subtle cognitive deficits later in life in spite of surgery, but also by better cognitive outcomes when earlier (< 6 months) and more immediately corrective (i.e. whole vault) techniques are utilized to intervene. This current study demonstrates that, compared to the numerous brain irregularities that are observed in infancy, there is substantially normalized brain growth morphology by adolescence when early comprehensive intervention occurs around the age of 7 months. This lends moderate evidence to the role of surgery in allowing for more normalized brain growth. Still, as also in line with the fact that cognitive deficits persist in spite of surgery, we also found that some brain shape irregularities do persist in spite of treatment, highlighting again how lasting effects can occur when an insult happens during a critical time in early development. Our study does not address correlation to neurocognitive data, and it relies on a small sample size. There may exist sub-populations of sNSC patients with more pronounced morphology differences. If so, it will be important to reveal how the degree of normalization achieved through surgery relates to baseline indices of head shape deformity and to cognitive outcomes. It will also be important to compare these adolescent DBM data to those obtained from infants prior to treatment. However, given the paucity of data on this important clinical topic and the novelty of our results, we believe our study provides encouraging data for the continued investigation of the effects of surgery on brain structure and function through MR analytics. We now show, for the first time, that the basic and important goal of allowing for more normalized brain growth is, at least in part, achieved through surgery.

Conclusion

Morphometry analysis using automated and objective MR DBM, demonstrates that sNSC children achieve substantial brain normalization following surgery compared to the numerous brain shape irregularities reported in previous studies on untreated sNSC infants. Allowing for brain normalization, since it is related to cognitive function, is a primary goal of calvarial release. Thus, these results support the early intervention through comprehensive cranioplasty to allow for more normalized brain growth in sNSC children.